However big all this news has been in the United States, I think it is dwarfed by the week's events in the world of football, that is, the kind without shoulder pads and helmets. As I write, England awaits The Derby on Monday, which looks to decide the English Premiership Championship. Last weekend, Real Madrid beat Barcelona in a Spanish La Liga match that probably determined their league's championship. Then, both teams went on to be upset in agonizing fashion in their respective Champion's League semifinals.
To one young man in my household, who's a big fan of Barca's Lionel Messi, it was a tough week.
I'm finally getting a chance to return to this blog. All this football (er, soccer) has me thinking about an earlier blog post where I began a discussion about sudden cardiac death in young athletes, and the potential for screening to prevent this tragedy. Thankfully, this was not a week in the sporting world with catastrophic outcomes like the sudden cardiac death of Piermario Morosini.
The issue of screening young athletes for conditions which could predispose them to sudden cardiac death (SCD) is a contentious one. I'm from Western Michigan, and just this spring near my hometown Wes Leonard collapsed and died after a high school basketball game. Deaths like these are always devastating. When I read these stories--I suspect when anyone reads these stories-- there is a natural desire to do something to prevent this from happening again. Just this last week, a good friend of mine from grade school and high school was asking me what I thought about his request that a local doctor perform a pre-season electrocardiogram (a.k.a. 'EKG' or 'ECG') on his daughter.
So. let's discuss the use of EKG's in the screening of young athletes in this country to prevent SCD. This will require some discussion of statistics.
First, it is important to define the population one wants to be screened. In general, athletes 35 years and older incur different risks than younger athletes. The vast majority of middle aged and older athletes who die in the middle of a contest have had a myocardial infarction or "M.I.". Screening this population is actually a fairly well defined proposition, and can often involve treadmill testing, looking at risk factors such as blood pressure, high cholesterol, and the like.
The question of how to screen younger athletes such as Morosini or Leonard is a different one, as the causes of their SCD are rarely from an M.I. Younger athletes who suffer SCD typically have a different set of predisposing risk factors, with esoteric names such as 'hypertrophic cardiomyopathy,' 'long QT syndrome,' "arrhythmogenic right ventricular dysplasia," and "Marfan syndrome." Reggie Lewis, the Boston Celtic, died of hypertrophic cardiomyopathy, and Flo Hyman, the Olympic Volleyball player, died of Marfan syndrome for instance.
We'll attempt to screen younger athletes then for this constellation of illnesses. Does a screening test exist? As things currently stand in the United States, whether a family is aware of it or not, the annual ritual of the preparticipation physical examination (PPE) is, among other things, a screening test for conditions that could cause SCD. And it is a very poor one with a sensitivity of about 2.5%. We'll discuss sensitivity in a moment, but for now it is sufficient to say that that makes the PPE a very poor screening test. There are no hard numbers, but one might begin to call a test 'good' when the sensitivity might be over 90%; so 2.5% is really abysmal.
Two other tests have been looked at in some depth, including echocardiograms (an ultrasound of the heart) and the EKG. Of the two, there exist much more data looking at the use of EKGs in screening for SCD. For instance, in Italy a group led by Domenico Corrado has been screening competitive athletes with EKGs since the 1970s. In part II of this post, which I'll be putting up on the blog later this weekend, we'll be looking in greater depth at screening EKGs. Before I close this post, a brief discussion of sensitivity and specificity, parameters that describe the accuracy and reliability of a screening test, is in order.
Sensitivity and Specificity are statistical measures of the performance of any test. There is a formal definition for each parameter, and a way of calculating these measures for tests ranging from the serum cholesterol test to the mammogram to the EKG. A good way of thinking of these parameters conceptually however uses the mnemonics 'SNout' for seNsitivity and 'SPin' for sPecificity. If you get a Negative result on a test with high seNsitivity, you can be reasonably assured you have 'ruled out' the disease you are screening for (hence the 'out' in 'SNout'). Conversely, if you get a Positive result on a test with high sPecificity you can be reasonably assured you have 'ruled in' (hence the SPin) the disease you are screening for.
An ideal test will maximize both specificity and sensitivity, but you can't have your cake and eat it too in the world of statistics: if you want to improve the specificity of a test, you will tend to reduce its sensitivity, and vice versa. There will always be a trade off between specificity and sensitivity. Issues of civil liberty aside, we have decided in the United States to use a low specificity way of screening for airline terrorists (we make everyone go through the scanner) in an attempt to keep sensitivity rather high (we want to make sure we don't miss the one terrorist who might get on the plane). Many screening tests will make this implicit trade off: they will maximize sensitivity at the cost of lower specificity, to ensure they don't miss a disease. The implicit cost of this method is to ensure a decent rate of false positives: the strange finding seen on mammogram that ultimately needs a biopsy to verify it is benign, or 'ok'; the disgruntled, aggrieved families who wonder why on earth their sweet grandmother had been patted down by the TSA. These are cases of false positives, consequences in some sense of the high sensitivity of the screening test involved.
There is one more concept to introduce now, before closing for the moment. The ultimate predictive value of a medical screening test is influenced by one more parameter: the underlying prevalence of the disease being screened for in the population. Prevalence is the frequency of a given disease in the population. For instance, we can say (sadly) that the current prevalence of obesity in the United States is 35.7%: more than a third of the adults in the U.S. are obese!!!
Getting back to the connection between prevalence and the testing performance parameters of sensitivity and specificty: for the statisticians out there, Bayes' Theorem is at play here. The general principle is that when screening for a rare disease, even tests with high specificty and sensitivity will have fairly poor predictive value. Put more simply, using even a decent test to look for a needle in a haystack will give you crappy results. What makes a test 'good' is not just intrinsic to the test itself (e.g. its specificity and sensitivity); it is dependent on the choice of population in which to use it. Put simply, one can potentially make even a 'good' test bad by choosing to use it in a low prevalence population. In many ways, that is the crux of the problem for the issue at hand.
We'll close for now, but in the next post we'll begin to look at the sensitivity and specificity of EKGs for screening for the conditions underlying SCD, and the prevalence of those conditions in the young athletic population. And then we'll begin to see whether it makes any sense to use screening EKGs to try to prevent these horrible outcomes.
No comments:
Post a Comment